We read with interest the VA CARDS (Veterans Affairs Coronary Artery Revascularization in Diabetes) study (1) comparing percu-taneous coronary intervention (PCI) and coronary artery bypass grafting (CABG) in subjects with diabetes. In our opinion, the most notable finding of the study is not the outcome of the trial, but the actual study design itself, that is, the stringent angiography-and clinically based inclusion and exclusion criteria, which makes application of the study's findings to real-world contemporary clinical practice highly questionable. Notably, of the 6,678 diabetic patients screened for this study, a staggering 6,080/6,678 (91%) of screened subjects did not meet angiographic requirements for the study, and only 198 subjects (3%) were randomly assigned to either CABG (n ϭ 97, 1.5%) or PCI with drug-eluting stents (DESs) (n ϭ 101, 1.5%) and completed the 2-year follow-up. Such a trial design is reminiscent of the PCI versus CABG randomized trials undertaken before the SYNTAX (Synergy Between PCI With Taxus and Cardiac Surgery) trial (2,3), where " cherry-picking " of patients before randomization was a major confounding issue (2% to 12% of screened patients were randomized in most trials) that effectively invalidated the results of these trials (4,5). The SYNTAX trial was designed to overcome these limitations by incorporating an all-comers design in which practically no patient was refused entry, with subjects either randomized (if determined by the Heart Team to achieve " equiv-alent anatomical revascularization " between CABG and PCI) or nested in registries (2,3). This was at the insistence of 7 cardiac surgeons (dubbed the " magnificent 7 ") during the design of SYNTAX to prevent selection bias, a view that was fully endorsed by the clinical and interventional cardiologists at the time. To give an example of the potential dangers of using highly selected populations in a clinical trial design, a recent meta-analysis of randomized trials undertaken before SYNTAX comparing PCI with CABG (6) showed CABG to be favored in older subjects and PCI to be favored in younger subjects, findings that have since been directly contradicted by the all-comers SYNTAX trial (where the opposite was shown) (7). In addition, the analyses demonstrat-ing the anatomic SYNTAX score in the VA CARDS study not to show any treatment effect between CABG and PCI warrant specific mention in that they were severely underpowered to draw any conclusions, even if considered hypothesis generating (8). Specifically, the majority of subjects in the VA CARDS study had low (Յ22) (CABG: n ϭ 47 vs. PCI: n ϭ 59) SYNTAX scores, with few subjects in the intermediate (23 to 32) (CABG: n ϭ 33 vs. PCI: n ϭ 24) or high (Ն33) (CABG: n ϭ 13 vs. PCI: n ϭ 12) SYNTAX scores, presumably due to the overwhelmingly restric-tive angiographic inclusion and exclusion criteria of the study as described earlier, thus making any comparisons of the low with the higher SYNTAX score tertiles practically meaningless. Even within the FREEDOM (Future Revascularization Evaluation in Patients With Diabetes Mellitus: Optimal Management of Mul-tivessel Disease) trial, (9) analyses based on the SYNTAX score tertiles appear to have been limited by power and were contra-dicted by those reported in the pre-stratified and powered diabetic subgroup of SYNTAX (10). The real lesson of the VA CARDS study is that randomization of subjects in a clinical trial is not enough and that an all-comers clinical trial design is warranted. Anything less will take us back to the confusing era of randomized trials performed before SYNTAX and will serve to cloud the medical literature.
Farooq, V., & Serruys, P. W. (2013). “Cherry-Picking” Patients for Randomized, Controlled Trials—Reliving the Past…. Journal of the American College of Cardiology, 61(24), 2492. https://doi.org/10.1016/j.jacc.2013.03.018